ONNX
Empowering the machine learning community with open-source interoperability through ONNX's common model representation and exchange format.
ONNX is an open-source machine learning model exchange format developed by the Open Neuromatch Foundation and contributors, enabling models to be transferred between deep learning frameworks with ease. Their mission is to promote openness, collaboration, and innovation in machine learning through a common representation for models and tools for conversion between frameworks. The ONNX project includes the ONNX format specification, conversion tools, community initiatives, documentation, and software releases under the Apache 2.0 license. Their goal is to reduce development efforts by facilitating model compatibility across deep learning frameworks.
About ONNX
ONNX is an open-source machine learning model exchange format, developed by the Open Neuromatch Foundation and a community of contributors. Their mission is to enable models to be easily moved between various deep learning frameworks and tools, thereby reducing the effort required in machine learning projects.
ONNX provides a common representation for machine learning models, allowing users to train models using one deep learning framework and then use them with another. They offer an extensive set of operators for handling various machine learning tasks, enabling developers to extend the format as needed.
ONNX's core values include openness, collaboration, and innovation. Their services are based on the principles of community-driven development and ensuring that their technology is accessible to everyone. Notable achievements include partnerships with major organizations such as Microsoft and Google Brain, demonstrating the widespread adoption and support for ONNX in the machine learning ecosystem.
The company's products and projects consist of:
- The ONNX format specification, which defines a common representation for machine learning models.
- A set of tools, including the ONNX converter, which enables users to convert models between deep learning frameworks.
- Regular community meetings and working groups, encouraging collaboration and idea sharing among contributors.
- Documentation and tutorials, providing resources for developers looking to get started with ONNX.
ONNX's software is released under the Apache 2.0 license and can be installed from PyPI or built from source using various operating systems such as CentOS, RHEL, and Fedora. They also maintain a code style guide (black) and use continuous integration pipelines for testing and ensuring high-quality releases.